Image by Google Images
“Obviously Nutanix is very influential among the first movers in this space, but it is almost in the same way that Netscape was influential in the browser and web server space.”
– Todd Brannon, Dir Product Marketing for Cisco UCS (TheNextPlatform 03/01/16)
You can’t say that Cisco lacks for confidence. Despite being the last server manufacturer to embrace hyperconvergence, the company didn’t hesitate to take shots at industry leader, Nutanix, while also making brazen claims about its brand new HyperFlex offering:
According to Cisco, only the UCS fabric interconnect architecture enables all traffic, even from the UCS blade servers, to reach any node in the cluster with a single hop. Perhaps this is why Cisco caps its HyperFlex clusters at eight nodes with only 4 clusters per UCS domain (or 32 nodes as opposed to a non-HX UCS domain of up to 160 nodes).
But Cisco, despite the success of UCS in the datacenter (which itself is driven by networking synergies), remains a networking company. Cisco must continue to sell a whole lot of switching and routing hardware. Even its SDN solution is designed to sell more Nexus switch ports.
The fallacy in Cisco’s argument is that if the network is such a barrier to HCI adoption, why has hyperconverged momentum been so extraordinary? Nutanix started shipping in late 2011 and since then, the industry has embraced HCI running on all kinds of hardware and network technologies.
More importantly, Nutanix’s version of HCI does not increase complexity but instead reduces it by adding intelligence to how and when the data touches the wire. Reliability is improved while bandwidth bottlenecks are slashed.
Data locality is one of the unique attributes of Nutanix web-scale. Data locality always keeps the data close to the VM which contains virtually all network within the server using the local virtual switches. Data locality eliminates the additional HCI stress on the network during steady state – which means most of the time.
The Nutanix software sits inside a VM and is a consumer of the hypervisor networking. It uses the existing virtual switch setup and configuration – all of which can be modified through Nutanix Prism. This can be a VMware Standard Switch, Distributed Switch or even a Cisco Nexus 1000V – and Nutanix does the same on Hyper-V.
The only time data leaves the node is if it is replicated to another node. Since Nutanix only uses one Controller VM (CVM) per host, only a basic 10G network is required. Any modern 10Gbps network already offers more than enough networking resources for a Nutanix cluster.
In terms of inter-VM communication, Nutanix nodes are no different than a UCS server running a hypervisor. If the hypervisor supports SDN, so does Nutanix. An SDN solution orchestrating networking functions will continue working irrespective of the HCI platform.
The Real Networking Challenge
Cisco says that, “Networking in most hyperconverged environments is an afterthought. With Cisco HyperFlex Systems, it is an integral and essential part of the system.” Central to this assertion is that enterprises still operate largely in silos. When compute or storage additions are required, only Cisco has the capability to seamlessly cross those silos.
Despite Cisco’s rant, it’s just not reality. There is no type of difficulty in four, eight or fifteen CVMs communicating with each other. But there is one networking challenge that is very real.
New networking products eventually cause redesign. As an example, the Nexus switch product line displaced the Catalyst switch product line via wholesale network redesign and forklift upgrades. For the most part, however, networking hardware lasts a long time. Organizations are often reluctant to throw away their investments.
Nutanix HCI significantly simplifies the networking design. The virtual switch is comparable to traditional top-of-rack. Top-of-rack is comparable to core switches. The networks are simple, clean and have a predictable traffic flow as they grow.
The networking challenge for organizations considering HCI is whether or not to continue utilizing a complex network architecture along with the IT functional silos supporting it. While Nutanix web-scale certainly doesn’t prohibit this approach, it does inevitably spur discussions about migrating to a much simpler and more efficient operation.
It’s All about Simplicity
Cisco markets Hyperflex as a single hyperconverged offering, but it’s really a collection of three existing solutions: Springpath software, network, and Cisco UCS. UCS was a very innovative product, as I wrote about in 2009, when it first shipped just over seven years ago. But UCS was designed for a hardware-based world, and that day has passed as customers have made it clear they want a software-defined datacenter.
A key tenant of the software-defined datacenter is simplicity. UCS is complex to install and time-consuming to upgrade. It includes custom ASICs, FCoE, Services Profiles and Templates, and Nexus switch integration – attributes that were tremendous seven years ago, but which merely add unnecessary complexity today. UCSM is a separate management console that provides more opportunities for bugs and outages.
Is HyperFlex “The Next Generation of HCI”?
Despite the incredible mindshare and momentum of HCI, Cisco appears to consider it only a niche offering. Fortunately for datacenter customers across the globe, its HyperFlex is going to change all that: “HyperFlex is, “The next generation of hyperconverged infrastructure…Cisco intends to take hyperconvergence mainstream in the Enterprise and accelerate it.”
If this grandiose declaration sounds familiar, it’s because Cisco made similar claims for prior introductions of new products:
ACE: “Next-generation Application Delivery Solution”
VXI: “Cisco® Virtualization Experience Infrastructure (VXI) accelerates Mainstream virtual desktop adoption…[and] delivers the next generation virtual workspace.”
Invicta: “Deploy next-generation integrated infrastructure”
I know very little about Springpath. It might well be fine HCI software, but clearly it is still very immature and untested. Even the Whiptail underpinning of Invicta was far more established.
HCI is much more critical to an organization than either virtualization hosts or storage. If the hyperconverged environment goes down, IT comes to a standstill. As independently validated both by achieving a Net Promoter Score (NPS) of 92 and by winning the Northface Scoreboard Award for World Class Excellence in Customer Service the last three years in a row, Nutanix long ago solved the reliability challenges of HCI.
Cisco’s comparison of Nutanix with Netscape implies that, like Netscape’s fate vis-à-vis Microsoft, the much larger Cisco will trounce Nutanix now that it has entered the HCI space. The analogy falls apart, however, in corporate vision. Netscape remained focused on its browser origins even as the Internet landscape dramatically changed whereas Nutanix has embraced enterprise cloud.
Maybe HyperFlex will end up being, as Cisco claims, the next generation of HCI. But Nutanix has already evolved well beyond hyperconvergence. HCI has become table stakes in the game of enterprise cloud. HCI, hypervisor and network are not end goals but simply checklist items in the requirement for efficiency and simplicity.
Nutanix is not burdened with the complexity of legacy network designs. In conjunction with its rapidly growing ecosystem, Nutanix focuses on enhancing the application lifecycle management experience from deployment to performance to management. This is how Nutanix is increasingly making infrastructure invisible.
Thanks to @evolvingneurons, @NutanixChris, Matt Northam, Thenu Kittappa and @vpai for edits and suggestions.
Fight the Fud – Cisco “My VSA is Better than your VSA”. 03/15/2016. Josh Odgers. CloudXC.
And for perspective…
Why Nutanix isn’t Singing the VSPEX Blues. 02/03/2015. Steve Kaplan. ChannelDisrupt.